Filling the gaps in the global prevalence map of clinical antimicrobial resistance

Surveillance is critical in containing globally increasing antimicrobial resistance (AMR). Affordable methodologies to prioritize AMR surveillance efforts are urgently needed, especially in low- and middle-income countries (LMIC5), where resources are limited. While socioeconomic characteristics correlate with clinical AMR prevalence, this correlation has not yet been used to estimate AMR prevalence in countries lacking surveillance. We captured the statistical relationship between AMR prevalence and socioeconomic characteristics in a suite of beta-binomial principal component regression models for nine pathogens resistant to 19 (classes of) antibiotics. Prevalence data from ResistanceMap were combined with socioeconomic profiles constructed from 5,595 World Bank indicators. Cross-validated models were used to estimate clinical AMR prevalence and temporal trends for countries lacking data. Our approach provides robust estimates of clinical AMR prevalence in LMIC5 for most priority pathogens (cross-validated q(2) > 0.78 for six out of nine pathogens). By supplementing surveillance data, 87% of all countries worldwide, which represent 99% of the global population, are now informed. Depending on priority pathogen, our estimates benefit 2.1 to 4.9 billion people living in countries with currently insufficient diagnostic capacity. By estimating AMR prevalence worldwide, our approach allows for a data-driven prioritization of surveillance efforts. For carbapenem-resistant Acinetobacter baumannii and third-generation cephalosporin-resistant Escherichia coli, specific countries of interest are located in the Middle East, based on the magnitude of estimates; sub-Saharan Africa, based on the relative prevalence increase over 1998 to 2017; and the Pacific Islands, based on improving overall model coverage and performance.

Automated summary

This article discusses a method to estimate antimicrobial resistance prevalence in countries lacking surveillance data by capturing the statistical relationship between socioeconomic characteristics and AMR prevalence. By establishing statistical models, accurate estimates of clinical AMR prevalence can be provided for low- and middle-income countries, informing 87% of all countries worldwide.